If you’re a scientist, stay tuned.
Everybody else change the channel. Alright so this one is for you guys
that are into gas exploration and weather forecasting
and compiling code and predicting which way
the stock market goes using ridiculous programs
and APIs using CUDA. This is the P7P55 Workstation
SuperComputer Motherboard, and as you can tell, it is meant
for supercomputers, as in Tesla-powered,
CUDA-taking advantage of, systems that are just pretty much
meant for scientists. So if you’re into gaming,
it does do SLI and CrossFireX, it does support P55 chipset,
up to 16 gigabytes of DDR3, up to 2133 megahertz
but basically again this is just for you guys
out there. What makes this board special though
is how many PCI Express slots it has, and ask you can tell, look at this.
We have five PCI Express x16 slots, so it’s very very impressive.
And if you guys have read up on P55, you know that it’s
a little bit different. It actually works on a different
architecture. The way that works is instead of like
X58 had the QPI, they got rid of that. They got
something called the DMI, Direct Media Interface, which means
that your PCI Express connectivity goes straight to the CPU.
And it also means that your memory goes directly to the CPU. Now they did
get rid of the triple channel memory but it still does talk directly
to the integrated memory controller inside the processor. The chip down here,
this is actually the P55 chip itself, and that’s a new type of chip.
It’s called the Platform Controller Hub and basically that controls
everything else on the board, your ethernets, your USBs,
your audio, everything else is going to go through this. Now how did they get
a chip that only supports 16 lanes of PCI Express to have 5 PCI Express
x16 slots? Like how is that going to work? I mean is it going to be like
x1, x1, x1 and then x8, x8? That wouldn’t be very much fun.
I’ll tell you how they did it. Right here, see this chip right here,
looks like the northbridge? Well there is no more northbridge
on P55, it’s a two chip solution. There’s this chip and there’s
this chip and that’s it. So you know what this is?
They added a third chip. It’s the NF200 from Nvidia.
You guys have seen it before on Skull Trails and gaming
boards and all this other stuff. Basically turns 16 lanes of PCI Express
into 32 lanes of PCI Express, in turn doubling it. Now on top of that
there’s still more connectivity. There’s more PCI Express connectivity
on this board and that’s because they’ve figured out a way
to bridge it over to P55 and use P55’s bandwidth to send
even more PCI Express connectivity so what basically is going
to happen on this board is people are going to be
loading it up with Tesla cards, maybe three of four Teslas
and they’re going to use a Quadro to get their graphics output out
of the computer into their monitor, so that’s what this board
is basically designed for. Now let’s talk a little bit about
the PCI Express compatibility. Check this out because it’s a little bit
complicated so stay with me here. Take a look at this first lane right here.
If you put one slot in there, one card, it’s going to be x16, right?
Now if you put all four, normally, that’s 32 lanes
of PCI Express, electronic lanes, and this can only do 16.
The NF200 chip turns that into 32, so if you populate four Teslas
in here they’re going to operate at x8, x8, x8 and x8. Now that’s
very very impressive. You might be thinking, “Well that’s great.
It’s PCI Express 2.0.” “It’s just as fast as the old
x16 PCI Express.” “That means I can do
4 way SLI on this board.” But that’s where you’d be wrong.
The problem with that is that although you have 32 lanes, you’re actually
only using 16 lanes worth of bandwidth. The chip is bifurcating that.
It’s actually the opposite of that, it’s spreading that PCI Express out
so you’re actually still limited in the bandwidth going
into the chip through the DMI. So that wouldn’t actually help
for that type of system but it would help in a Tesla-based system,
so this really is really, really good for a workstation with multiple Teslas.
It’s not as great for an SLI or Crossfire setup but it still works.
It still works with SLI and Crossfire, especially with two-way SLI,
it would work really good. Three-way SLI, that third card, not all
of its power may get to the processor, same way with three-way Crossfire.
So it still works, you can still use it for gaming.
You can still use it for overclocking, but again, the board is designed
specifically for workstations for using CUDA-based, Tesla cards,
the C1060 or whatnot and some Quadro cards or if you need
to do a lot of big, very large monitors, things like that. It is called
the supercomputer for a reason, and that’s because you can get
like 4 teraflops of computing power on a computer that fits
inside of a case, versus having to rent time
at a huge array somewhere, some supercomputer that’s going
to cost you millions of dollars. You’ve got to get on a list and wait.
This way scientists can get to work at their desks. They don’t
have to wait for their time to come up on the supercomputer that’s
really busy, everybody wants to use it. They get to use it right away.
Now there’s a lot of other really, really cool features on this board,
let me show them to you. I’m going to start off
with the back panel, I want to show you what kind
of connectivity on here. You get two PS2s right there,
you have 8 USBs 2.0s on the back here in total,
so that’s a lot. You get dual Gigabit ethernet,
you also have a Firewire port right there, S/PDIF in both coaxial and the optical
version right here which is the digital. You get 7.1 channel
HD audio right there, so if you want to listen